Development of effective combination therapy for cancer is challenging because many cancer drugs act on intersecting signaling pathways that can interfere with each other. For example, it is well established that drug-drug interactivity can change drastically, e.g., from synergy to antagonism, depending on the treatment conditions (drug concentrations, treatment time, and sequencing of the drugs). Further, under in vivo conditions, drug concentrations change with time (i.e., pharmacokinetics or PK) and different drugs have different PK, which make it difficult to translate the findings in cultured cells where drug concentrations are typically kept constant. We propose to develop multiscale, generalizable computational PK and pharmacodynamic (PD) models to address these challenges. First, we will develop predictive in vitro PD models for single agents. These PD models employ a combination of deterministic models (that designate the fate of a single cell based on drug actions and cell cycle location) and probabilistic models (that determine the fate of all cells). These models jointly depict the response of an individual cell and the overall response of whole cell population (as the collective response of individual cells), as mathematical functions of a treatment (drug concentrations, treatment time) and chemosensitivity of a cell. Second, we will develop predictive in vitro PD models for combination therapies. Drugs can interact on two levels, i.e., cell cycle distribution (cell cycle interactivity) and molecular targets (molecular interactivity). We will extend the above approaches for single agents to develop two-drug-combination models for three types of combinations: (a) drugs with only cell cycle interactivity, (b) drugs with molecular interactivity where both drugs have cytotoxic effects, and (c) drugs with molecular interactivity where one drug does not have cytotoxicity on its own but can enhance and reduce the activity of the other drug. Third, we will develop methods to convert in vitro PD to in vivo PD. We will address two issues, i.e., conversion of CxT under in vitro conditions (constant C) to in vivo situations (changing C), and extend the in vitro PD models to include the non-cycling G0 cells present in vivo tumors. Lastly, we will integrate in vitro PD models with in vivo PK models and evaluate the performance of the integrated models. We will develop integrated PK-PD models to describe the in vivo effects of single agents, followed by models for their combinations. Model performance is evaluated in tumor-bearing animals. The proposed models are first-of-its-kind and will enable the computation of outcomes of potential combinations of different drugs and/or different in vivo treatment schedules/sequences. Such predictive models can reduce the uncertainty in outcomes and the amount of experimentation and thereby accelerate the development of effective combination cancer therapies.